My currents research interests are a diverse set near the intersection of Machine Learning and Computer Vision. I work on stochastic optimization, deep learning, and computational biomolecular imaging. In the past I've worked on learning for autonomous robotics, unmanned aviation, and large scale image retrieval. I am also interested in social entrepreneurship and technological approaches to social change.
Since 2015 I've focused on algorithms and computational techniques for Electron Cryomicroscopy (cryo-EM). Cryo-EM is a revolutionary new tool that enables direct discovery of the 3D structure of protein molecules. This ability is fundamentally game-changing for drug design and understanding the inner workings of biological life. A major portion of the cryo-EM process is computational, where we use large scale inference and optimization to resolve atomic-level details of molecules from terabytes of electron microscopy data. Read more about cryo-EM at Nature. On the right is an example cryo-EM structure resolved by algorithms I work on.
cryoSPARC: Algorithms for rapid unsupervised cryo-EM structure determination
Ali Punjani, John Rubinstein, David Fleet, Marcus Brubaker
Nature Methods 14, 290-296 (2017)
[paper]
[readcube]
[software]
[blog]
[F1000]
The algorithms and software developed in this work are now being expanded and commercialized by
Structura Biotechnology Inc. You can read more about this work in the following news articles:
[UofT News]
[Varsity News]
Algorithms for Reducing the Computational Burden of Cryo-EM (Invited Talk)
Outlier Rejection Methods and Branch-and-Bound Image Alignment in CryoSPARC (Invited Poster)
Ali Punjani, John Rubinstein, David Fleet, Marcus Brubaker
Gordon Research Conference: 3D Electron Microscopy 2016
Building Proteins in a Day: Efficient 3D Molecular Structure Estimation with Electron Cryomicroscopy
Ali Punjani, Marcus Brubaker, David Fleet
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 2016 (in press)
Alpha-Beta Divergences Discover Micro and Macro Structures in Data
Karthik Narayan, Ali Punjani, Pieter Abbeel
International Conference on Machine Learning (ICML) 2015
Building Proteins in a Day: Efficient 3D Molecular Reconstruction (Oral Presentation)
Marcus Brubaker, Ali Punjani, David Fleet
IEEE Computer Vision and Pattern Recognition (CVPR) 2015 [arXiv preprint]
Featured in MIT Tech Review!
Deep Learning Helicopter Dynamics Models
Ali Punjani, Pieter Abbeel
IEEE International Conference on Robotics and Automation (ICRA) 2015 [pdf]
[slides]
[poster]
Microscopic Advances with Large-Scale Learning: Stochastic Optimization for Cryo-EM (Oral Presentation)
Ali Punjani, Marcus Brubaker
Neural Information Proc. Sys. (NIPS) 2014 Workshop on Machine Learning in Computational Biology [slides]
Published on arXiv [pdf]
Machine Learning for Helicopter Dynamics Models
Ali Punjani, Pieter Abbeel
MSc Thesis, UC Berkeley. Committee Members: Pieter Abbeel, Jitendra Malik, Ken Goldberg [pdf]
Fast Exact Search in Hamming Space with Multi-Index Hashing
Mohammad Norouzi, Ali Punjani, David J. Fleet
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 36, no. 6, 2014 [preprint]
Fast Exact Search in Hamming Space with Multi-Index Hashing
Mohammad Norouzi, Ali Punjani, David J. Fleet
IEEE Computer Vision and Pattern Recognition (CVPR), 2012 [pdf]
Research Engineer and Data Scientist - PrecisionHawk Inc. - [Behind the Scenes Video]
CSC2515 - Graduate Introduction to Machine Learning - Winter 2015
NSERC Canada Graduate Scholarship (Doctoral) - 2015
NSERC Canada Graduate Scholarship (Masters) - 2013
I work on various projects, most as hobbies. I am involved in CNC Milling, 3D Printer design, digital artwork, arabic calligraphy, model aircraft design, etc. I completed my undergraduate degree as an engineer (Engineering Science Aerospace 1T2) and I often like to build things with my hands using tools other than a keyboard and mouse.